Comparación entre regresión logística y redes neuronales para predecir cáncer de piel en perros

• Autor: Chávez Martínez, Renato
• Formato: tesis de grado
• Fecha de Publicación: 2019
• Institución: Universidad de Lima
• Repositorio: ULIMA-Institucional
• Lenguaje: español
• OAI Identifier: oai:repositorio.ulima.edu.pe:20.500.12724/8401
• Enlace del recurso: https://hdl.handle.net/20.500.12724/8401; http://doi.org/10.26439/ulima.tesis/8401
• Nivel de acceso: acceso abierto
• Materia: Cáncer; Perros; Prospectiva; Cancer; Forecasting; Dogs; https://purl.org/pe-repo/ocde/ford#2.02.04

To ascertain if a dog has the predisposition to develop skin cancer is a challenge for both veterinarians and pet owners. Logistic regression models and neural networks have been used widely in the field of human medicine to make predictions; the present study approaches the comparison between these two technics to predict skin cancer in dogs. The variables we analyzed were age, sex, breed, sun exposition, albinism and, dermatitis. These variables were validated by the correlation coefficient and the principal component analysis. The obtained results showed that the backpropagation neural network technique with a cross validation is better than the logistic regression. The neural network’s accuracy value was 89.6% while only 84% for the logistic regression.